ES
To obtain a better insight in the character of a systematic deformation it is
necessary to test individual groups of residuals or even single residuals. For
this purpose appropriate test criterions can also be derived as it was demonstra-
ted for additional parameters in section D1. But notice the difficulties arising
with high correlations with respect to the confidence interval setup.
It is an unpleasant experience that the inadequate presicion of control points
does often disturb the photogrammetrically obtained coordinates, thus leading
to wrong conclusions with respect to the photogrammetric accuracy potential. The
disturbing influence of control points may be caused randomly or systematically.
However, a statistical treatment of the residuals at control points is a helpful
device to arrive at better conclusions with respect to the control accuracy in
any case. {
Of course it is not easy or often impossible to seperate the photogrammetric
effect from the control influence and to seperate the random control errors from
the systematic control errors. But if a self-calibrating adjustment is performed
we can assume that the main photogrammetric systematic influence is compensated, (
so that the residuals at control points may indicate at least a part of the
original systematic control deformation.
A global test on the significance of control residuals can be based on
Bs pP tuas ; d
9 :5,d d,] d,
M
number of control (50)
coordinates
(Mostly dxP will be assumed as dxP = 0).
Hence we get the test criterion
= R
T = Ii , (51)
0
with
oF oP Tat opp
R s (dx -dxl ) Qpp (dX -dx" ) . (52)
Under H, T is distributed as the central F(d,r).
i
Usually the weight matrix Pp of the control point coordinates (see(1)) will be a
diagonal matrix. If the weights are not too small then Q5) - Q pp is the weight
coefficient matrix of the estimated control coordinates - is diagonal dominant,
i.e. the coefficient columns of the control residuals dx; are mostly orthogona]
to the other unknowns of system (1). p
So the application of the t-test for the testing öf individual residuals dx; with
à type I error size a:
15% ow] fa &j4 ; a = type I error size of (53)
the global test (50)
is more successfull than in the cases mentioned before.